Using AI at work without losing your focus

By Tophik · Updated 14 October 2026 · 6 min read

AI assistants can draft an email in seconds, summarise a long report and suggest ten ways to start a presentation. They can also turn a focused morning into an afternoon of half-read answers, rewritten prompts and tabs you never closed.

The difference is rarely the tool. It is how you use it. This guide brings together what the best research so far says about AI at work, and turns it into a simple routine that keeps you in charge.

In short

In large studies, AI assistants made people faster and better at many tasks, with the biggest gains for less experienced workers.
On tasks just outside what the AI can do, people using it got the answer wrong more often, while sounding more convincing.
Decide what you need before you open the tool; write a brief, not a wish.
Check every fact, figure and quote before it leaves your desk.
Don’t type anything into a public chatbot that you wouldn’t want made public; the UK’s National Cyber Security Centre gives the same advice.

In this guide

What the research says about AI at work
The jagged frontier: knowing when not to use it
A focused way to work with AI
Writing a brief, not a wish
Checking the output
What not to type into a chatbot
Protecting your attention
Questions people ask

1. What the research says about AI at work

Two studies are worth knowing, because they were large, careful and done with real workers rather than students in a lab.

Customer support: 15% more problems solved

Economists Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied more than 5,000 customer support agents at a software company as an AI assistant was rolled out. Published in the Quarterly Journal of Economics in 2025, they found that access to the assistant increased the number of issues resolved per hour by about 15% on average. The gains were largest for newer and less skilled agents; the most experienced agents saw small gains in speed and a small drop in quality.

Consultants: faster and better, until they weren’t

A team from Harvard Business School and other universities ran an experiment with 758 consultants at Boston Consulting Group. On a set of realistic tasks that the AI was good at, consultants using GPT-4 completed 12.2% more tasks, 25.1% faster, and their work was rated more than 40% higher in quality. But on one task designed to sit just outside the AI’s abilities, consultants using it were 19 percentage points less likely to reach the correct answer than those working without it.

Put together, the message is clear: AI can be a real help, and it can make confident mistakes that you then repeat. The skill is knowing which situation you’re in.

2. The jagged frontier: knowing when not to use it

The Harvard researchers called the boundary between what AI does well and badly a “jagged technological frontier”. Tasks that look equally hard to us can sit on opposite sides of it. In daily work, a rough guide:

Usually good at	Usually risky
First drafts of routine writing: emails, outlines, summaries	Exact facts, figures, dates and quotes
Rephrasing for a different reader or tone	Anything that depends on information it hasn’t seen, such as your company’s latest numbers
Brainstorming options and counter-arguments	Legal, medical or financial conclusions
Explaining a concept you can then check	Problems where a plausible answer and a correct answer look the same

The frontier moves as tools improve, so the best habit is to test new uses on low-stakes work first, where a mistake is cheap and easy to spot.

3. A focused way to work with AI

Most lost time with AI comes from opening it too early, before you know what you want. A simple routine:

Think first, for two minutes. Write down on paper what you need and what “good” looks like. If you can’t, the AI can’t either.
Batch your AI tasks. Collect them and do them together in one block, rather than switching to the chatbot every few minutes.
Give it a clear brief (see the next section) and ask for one thing at a time.
Timebox it. Give yourself a fixed time, say ten minutes, to get a usable draft. If it isn’t working, write it yourself.
Edit as the expert. Treat the output as a junior colleague’s draft: useful, but yours to correct and sign off.

Try this: pick one repetitive writing task you do every week. Write a reusable brief for it, save it in a note, and use the same brief each time.

4. Writing a brief, not a wish

“Write me a good email to a client” is a wish. A brief gives the tool what a colleague would need:

Goal: what the text needs to achieve. “Get the client to agree a new delivery date.”
Reader: who it is for and what they already know.
Material: the facts to use, pasted in (minus anything confidential).
Format: length, tone, structure. “Under 120 words, friendly, one clear question at the end.”
Limits: what not to do. “Don’t promise a refund. Don’t invent dates.”

Keep the briefs that work in a simple document. A small library of five good briefs saves more time than any clever trick.

5. Checking the output

AI tools write fluently whether they are right or wrong, which is exactly why the consultants in the Harvard study were caught out. Before anything leaves your desk:

Check every number, name, date and quote against the original source.
Ask: is anything here that I didn’t give it? New “facts” are the ones most likely to be invented.
Read it out loud once. Generic phrases and things you’d never say stand out immediately.
For anything important, ask a person who knows the subject.

6. What not to type into a chatbot

The UK’s National Cyber Security Centre advises not including sensitive information in queries to public AI tools, and not asking questions that would cause problems if they were made public. Queries are visible to the company running the service and may be stored. In practice, keep out:

Personal data about customers, colleagues or yourself (names with health, money or HR details).
Passwords, keys and access details, ever.
Confidential business information: contracts, unreleased numbers, plans.

Check your employer’s policy first. Many organisations provide an approved tool with stronger privacy terms; use that one for work.

7. Protecting your attention

AI tools are now built into email, documents and phones, which means more suggestions and pop-ups competing for your attention. Research on interruptions, such as the University of California, Irvine study we discuss in our Pomodoro guide, links frequent interruptions with more stress and frustration. A few settings help:

Turn off AI suggestions that appear while you type, if you find them distracting. Call the assistant when you want it.
Keep one chatbot tab, not five. Close it when the batch is done.
Do the hard thinking (the plan, the decision, the key paragraph) yourself first, then use AI to speed up the rest.

A calm, tidy desk with a closed laptop, keyboard, mouse and a mug — Think on paper first; open the tool when you know what you need.

Questions people ask

Will using AI make me worse at my job?

It depends on how you use it. Letting it think for you on things you should understand is risky; using it to speed up routine work you can check is generally sensible.

Should I tell people I used AI?

Follow your employer’s policy. In general, be open about it where it matters, for example when the reader relies on the work being yours or being accurate.

Which AI tool is best?

The tools change too quickly for a fixed answer. Use the one your organisation approves for work, and judge it on your own tasks.

Used well, AI is like a fast, tireless assistant who is sometimes confidently wrong. Keep the thinking, the checking and the final word for yourself.